The Provisions paper asks whether a Bitcoin exchange can demonstrate sufficient reserves for customer accounts without publishing its addresses, total holdings, liabilities or customer information. It develops a privacy-preserving proof system and evaluates computation and proof sizes, including large customer sets. The authors also discuss an extension intended to prevent exchanges from colluding to conceal shortfalls.

Read the proof statement and its inputs before interpreting the word solvency. A cryptographic system reasons about the accounts, assets and obligations represented in its model. It does not independently discover an omitted creditor, a contractual restriction or a false external record. The paper is a concrete research contribution, not a certification of any exchange that links to it. An implemented proof and the truthfulness and completeness of its input ledger are separate evidence requirements.

Buterin's 2022 technical essay explains Merkle sum trees and zero-knowledge approaches to exchange liability proofs. A customer's branch can establish that a particular balance contributes to a published root and total. The essay discusses privacy leakage and the need to handle negative balances correctly, because an invented negative account could otherwise understate an aggregate obligation.

A customer checking their own inclusion has learned something meaningful about that published snapshot. They have not necessarily verified that every other customer was included. Likewise, one root should not silently represent different snapshots for different audiences. A careful implementation review follows the balance encoding, non-negativity checks, snapshot time and distribution of the common commitment. These checks make the limited statement precise without inflating it into a complete financial audit.

Imagine a fictional service with customer balances of 20, 30 and 50 units. A proof correctly shows that those entries sum to 100. If a separate lender is owed another 40 units but never appears in the committed dataset, the arithmetic can remain perfectly correct while the service's total obligations are understated. This is an illustrative accounting boundary, not an accusation about a real provider.

Now give the service 110 units in visible wallets, of which 60 were borrowed for the snapshot. A proof of key control does not explain the corresponding debt or whether the assets are freely available for customer withdrawals. To evaluate the claim, align the asset and liability snapshots, identify the included legal entities and examine encumbrances. More sophisticated cryptography cannot supply facts that were excluded from the statement being proved.

The PCAOB Office of the Investor Advocate's March 2023 advisory warns that proof-of-reserve reports are not equivalent to audits. It highlights possible omissions involving liabilities, borrowed assets, customer rights, internal controls and changes after a point-in-time check. The advisory concerns the scope and limitations of such reports; it does not say that every cryptographic inclusion proof is useless.

For research, identify who performed which procedures, under which standard, at what date and over what population of records. Distinguish an agreed-upon-procedures report from an audit opinion and from a self-published cryptographic proof. Then describe what the evidence supports in plain language. An inclusion check can strengthen transparency, while a judgment about an entity's ability to meet obligations requires a broader body of evidence.